The Skill Supply Chain Problem

Welcome to Amaze Networks for Wednesday, April twenty-ninth. Three consecutive weeks of AI infrastructure supply chain attacks — and today we've reached the third layer: agent skill registries.

And the reason this one is different from the last two is that the attack vector isn't code at all. No malware, no exploit. Just a text file that tells an AI agent what to do — and the agent does it.

So the lead story is ClawHub. ClawHub is the skill marketplace for an agentic AI framework called OpenClaw. Thirty malicious skills were published there, accumulated about nine thousand eight hundred downloads, and silently turned all those agents into a crypto mining swarm.